Re: performance issue after upgrade to oracle linux 32 bit.

From: joel garry <>
Date: Wed, 4 Nov 2009 16:56:36 -0800 (PST)
Message-ID: <>

On Nov 4, 2:26 pm, Charles Hooper <> wrote:

> -
> Randolf: regarding the optimizer bug that was fixed in which
> was related to ROWNUM and incorrect cardinality estimates, I was
> unable to force to reproduce the incorrect cardinality
> estimates by changing the OPTIMIZER_FEATURES_ENABLE parameter to match
> a lower release number.  So, I guess that some fixes are buried deep
> in the code.  I even tried disabling some of the bug fixes without
> success.  This seems to imply that changing the
> OPTIMIZER_FEATURES_ENABLE to an older version may not always reproduce
> the old optimizer behavior, which is a bit unfortunate.

Yes, Jonathan blogged about an example of this in "FBI problem" ( ) But it's still usually worth a try, if just to narrow possibilities.

Also, just to clearly emphasize to the OP, it is recommended in a lot of places not to set OPTIMIZER_INDEX_COST_ADJ and OPTIMIZER_INDEX_CACHING, at least until testing shows they help - it's likely a mistake to keep the settings from earlier versions. See for some idea of what to look for.


-- is bogus.
Received on Wed Nov 04 2009 - 18:56:36 CST

Original text of this message